Revolutionizing Pancreatic Cancer Diagnosis with AI: Early Detection and Population Screening

2023-06-06 15:24:46

Pancreatic cancer is still detected very late on average, making it one of the most deadly cancers. An earlier discovery using a smart AI tool can advance the diagnosis years, allowing the disease to be treated at a much earlier stage. In general, early diagnosis of cancer can significantly increase treatment options and therefore survival.

Insulin

The pancreas (pancreas) is an elongated and important organ in the abdomen. The pancreas makes substances that are important for digestion. These are, for example, enzymes for digesting fat, but also hormones that are important for regulating the sugar level in the blood. The pancreas therefore plays a major role in diabetes. In type 1 diabetes, for example, the pancreas no longer produces insulin at all, because the cells that have to do this have been destroyed or shut down by its own immune system. Cancer of this important organ initially gives hardly any or vague complaints, which means that pancreatic cancer is only discovered at a late stage and is therefore usually fatal.

Population screening for pancreatic cancer?

Thanks to artificial intelligence, large-scale screening may become possible in the long term. In the Netherlands, certain target groups are already regularly checked for colon cancer and breast cancer. However, that is a costly and very time-intensive matter, which can become less complex through the use of AI. Radiologist Rozemarijn Vliegenthart of UMCG recently reported, for example, that population screening for lung cancer with AI will eventually become possible manageable and can be affordable. Thanks to the new AI tool from Harvard and the University of Copenhagen, population screening for pancreatic cancer may also come into the picture in the long term. This is because AI can view CT scans, but also other images, at lightning speed and accurately.

This mainly concerns the screening of people who have an increased risk of pancreatic cancer. Senior researcher Chris Sander, a faculty member in the Department of Systems Biology at HMS’s Blavatnik Institute, said: “One of the most important decisions clinicians face on a daily basis is who is at high risk for a disease and who would benefit from further testing. But those tests can lead to more invasive and expensive procedures that come with their own risks. An AI tool targeting those at highest risk of pancreatic cancer, who would benefit most from further testing, could go a long way in improving clinical decision-making.”

Innovative research

The necessary is also being done in the Netherlands innovative research done to pancreatic cancer and AI is also used. A concrete example of this can be found in recent research in Amsterdam UMC. This is not about predicting pancreatic cancer, but about more precise, tailor-made treatment. In this research, a team of doctors, researchers and data scientists, together with experts from software supplier SAS, are working on the development of an AI tool that objectively determines how a tumor responds to a treatment.

In the new Harvard study However, it is not about tailor-made treatment, but about the early detection of pancreatic cancer. For this, the AI ​​algorithm was trained on two separate datasets with a total of 9 million patient records from Denmark and the United States. The researchers “asked” the AI ​​model to look for telltale signs based on the data in patient records. Based on combinations of disease codes and when certain diseases started, the model was able to predict which patients would develop pancreatic cancer in the future.

AI tool is accurate

It was striking that a number of symptoms in particular, which are normally not directly related to pancreatic cancer, nevertheless had a certain predictive value. The researchers eventually tested different versions of the AI ​​models for their ability to detect people at increased risk of disease development.

Remarkably often it was possible to advance the diagnosis by up to two years. Overall, each version of the AI ​​tool was significantly more accurate at predicting who would develop pancreatic cancer than current estimates. The researchers even think that their model is at least as accurate in predicting the onset of disease as current genetic DNA tests; which, by the way, are only available to a small subgroup of patients.